Claude Opus 5, Now Live In Terminal X

Anthropic's Claude Opus 5 is now running in Terminal X. Anthropic positions it as an everyday model that comes close to the frontier intelligence of Claude Fable 5 at half the price, and reports state-of-the-art results on coding and knowledge-work evaluations. In our own grading of Opus 5 across live analyst queries, we saw a meaningful gain in reasoning and answer clarity in financial test sets covering filings comprehension and end-to-end analyst tasks.

The gains that matter to an investment team show up in numerical work and in long-running agentic runs. Our own internal testing is consistent with Anthropic's early-access finance testers: for finance workflows, we found a ~9 percentage point accuracy increase on average across effort levels, with a third fewer tool calls and 60% less time. In addition, Anthropic's applied AI team cites numerical reasoning, table work and sharper critical thinking on financial research workflows.

Independent evaluation points the same direction, with private finance benchmarks placing Opus 5 first on CorpFin v2 at 73.19% and first on Finance Agent v2 at 58.63%.

Terminal X is model-agnostic by design. Our Agent Orchestrator selects the optimal model and reasoning depth at each step of a workflow, so Opus 5 slots into the pipeline where its strengths compound: parsing a Private Data Room during diligence, reconciling projections against actuals in filings, holding a thesis together across a long multi-agent run, and producing the final deliverable at the end.

At Terminal X, we've found that a benchmark score isn't the last mile. Models alone can't produce industry-standard work, because no model knows which vendor feeds matter to your desk, how your team structures an IC memo, or where the real numbers live across twenty systems. Terminal X FDEs close that gap. If you're interested in integrating Terminal X across your firm's entire internal knowledge base, reach out to us for a pilot at [email protected].